Project AI Code Assistant (or similar name)

Intelligent Code Management Powered by AI

Project AI Code Assistant is a powerful Rust web application designed to help developers effortlessly understand, manage, and evolve their source code projects using cutting-edge AI. By converting your codebase into a structured YAML format and leveraging advanced vector embeddings and Large Language Models (LLMs), it provides unparalleled capabilities for code exploration, analysis, and modification.

Why Use Project AI Code Assistant?

In large codebases, finding relevant information, understanding complex interactions, and making informed changes can be a significant challenge. This tool solves that by:

• Breaking Down Complexity: Transforms raw code into a clean, structural YAML representation, making code intent clear.
• Unlocking Semantic Search: Go beyond keyword searches to find code based on what it does, not just what it says.
• Empowering AI Assistants: Chat with LLMs about your specific code, getting intelligent suggestions and analysis in real-time.
• Streamlining Code Updates: Iteratively refine code with AI guidance, always ensuring the LLM understands the current state of your files.

A Note on Development: This application has been leveraged by the author to develop and advance its own codebase since its earliest stages. Its growing utility has allowed the author to phase out reliance on external LLM web interfaces entirely. This self-reliance eliminates tedious manual steps and the context-switching headache of other tools, centralizing the entire development workflow—from intelligent code analysis to streamlined Git operations—all within a single, focused browser tab.

Features

• Automated YAML Generation: Converts your source code files into a structured, language-agnostic YAML format, focusing on code structure, documentation, and intent (no implementation details included!).

• Centralized YAML Management: Create, retrieve, update, and delete YAML representations for your projects via intuitive API endpoints.

• Intelligent Code Search (Semantic Search):

  • Leverages vector embeddings and a Qdrant vector database for highly accurate semantic search.
  • Find relevant code files across your project using natural language queries – discover code based on its purpose, not just keywords.

• AI-Powered Code Analysis & Chat:

  • Context-Aware LLM Interactions: Engage in interactive chats with LLMs (e.g., Anthropic, OpenAI, Gemini) to analyze and discuss your code.
  • Smart File Selection: Vector search results provide a starting point, and the LLM can further recommend additional relevant files.
  • Dynamic Context Control: You have full control over which files are shown to the LLM at any point in the conversation, ensuring the LLM always has the precise context you need.
  • Live Code Updates: The LLM is explicitly informed that the code context can be updated with each message, enabling seamless, iterative code modification discussions.
  • Syntax-Highlighted Responses: LLM code suggestions are automatically parsed from Markdown and rendered with beautiful syntax highlighting for optimal readability.

• Project & File Management:

  • Dedicated project pages to view generated YAML representations.
  • Selective regeneration of YAML for individual files.
  • Save and resume previous analysis chats and query histories.

• Integrated Git Workflow:

  • Streamlined Branch Management: Easily create new feature branches directly from your chat session to isolate development, allowing you to work on significant updates without affecting your main codebase.
  • Seamless Branch Switching: Use the built-in branch selector to view the current active branch and switch between branches within the application.
  • Flexible Commit Options:
    • Auto-Commit: Enable per-chat auto-commits to automatically save changes between user and AI messages to the active branch, associating conversation progress directly with code milestones.
    • Manual Commit: Take granular control by manually committing changes with a custom message using the dedicated "Commit" button.
  • Push to Remote: Synchronize your local commits with your remote repository (e.g., GitHub) at any time with a single click.
  • Effortless Merging: Merge your feature branch back into the main development branch once your changes are complete and reviewed.
  • Integrated Git Credentials: Configure your Git author name, email, and remote authentication credentials (like GitHub Personal Access Tokens for HTTPS) directly through the application's UI, ensuring smooth push operations.
  • Conflict Handling: In case of merge conflicts, the application will notify you, requiring manual resolution outside the app to maintain full control over your codebase.
  • AI's Role: The AI assistant provides code suggestions and insights, while all Git operations (branching, committing, merging, pushing) remain fully under user control.

• User-Friendly Web Interface: Access all features through a modern web application.

Installation

1. Clone the repository:

   git clone https://github.com/scionsamurai/code-to-yaml-vector-search.git
   cd code-to-yaml-vector-search

2. Start Qdrant Vector Database: Project AI Code Assistant uses Qdrant for blazing-fast vector search. The easiest way to run it is with Docker:

   docker run -p 6334:6334 -p 6333:6333 qdrant/qdrant

Usage

1. Configure API Keys:
   • Option 1 (Recommended): Create a .env file in the project root and add your API keys (e.g., OPENAI_API_KEY, GEMINI_API_KEY, ANTHROPIC_API_KEY). See the llm_api_access crate documentation for example .env structure.
   • Option 2: Add/update keys directly through the web interface (see step 4).
2. Run the Application:

   cargo run

3. Access the Web Interface: Open your browser and navigate to http://localhost:8080.
4. (Optional) Update Environment Variables: If you skipped step 1, click the "Update Environment Variables" button on the home page.
5. Create a New Project: Provide a project name, the path to your source code directory, and the programming languages used. The application will automatically generate YAML files and their embeddings.
6. Explore & Analyze:
   • View the generated YAML representations on the dedicated project page.
   • Use the semantic search bar to find relevant code files.
   • Initiate a code analysis chat to discuss and modify your codebase with the LLM.
   • Dynamically select which files are visible to the LLM within the chat interface.

Configuration

The application can be configured using environment variables or a .env file in the project's root directory.

• OPENAI_API_KEY: API key for OpenAI models (embeddings, chat).
• OPEN_AI_ORG: Organization ID for OpenAI API (if applicable).
• GEMINI_API_KEY: API key for Google Gemini models.
• ANTHROPIC_API_KEY: API key for Anthropic models (default for chat analysis).
• QDRANT_SERVER_URL: URL for the Qdrant vector database (defaults to http://localhost:6334).

License

This project is licensed under the MIT License.